Mapping Floods

Foster City presents an interesting case study on how to plan for the monetary losses of bay area flood risks. For the purpose of this analysis, I will look at nine scenarios of 3 possible sea levels in the bay and 3 possible storm intensities and their associated damage to vehicles within the region. The extremity of a storm is often described on a multi-annual timescale (i.e. 20 year storm, 100 year storm etc.), which Our Coast Our Future has modeled in combination with rising sea levels.

The following map gives a sense of the variation across different combinations of sea level rise (SLR) and storm events. Worth noting is that for every scenario involving 50cm of SLR, the entirety of Foster City is subjected to flooding.

The change to total submersion happens much more quickly than other regions in the area, and reflects the near instantaneous breaching of protective seawall. While there are plans for significant improvements to the levee around its perimeter, Foster City is at an enormous risk for flood damage once protective infrastructure fails.

To quantify one aspect of that damage, I will incorporate Census data on vehicle ownership within the city. I overlay the rasterized image file of each flood to generate an average depth intersecting with every building in Foster City. I am using OpenStreetMap building information, which is imperfect in documenting every type of building in this region, but works well in identifying which structures each flood interacts with. I want to extrapolate building damage to vehicle damage, however, so I need to associate a vehicle count with each building.

From the Census reported block group level data, Foster City has 37,289 vehicles in 2020. Using US Army Corps of Engineers damage data, I can relate flood depth to a percentage damage for a vehicle. It is worth noting that there is a count of 7,221 households with only one vehicle, and 718 without any. In other words, there are nearly 8,000 homes with little means of transportation in the case of extreme flooding. For the purpose of this analysis I will be using the data on sedan percent damage as an approximation for all vehicles in the study area. The damage percentage curve associated with flood depth can be visualized below at various SLR during a 100 year storm event.

To get an estimate of the number of vehicles for every home, I can connect building information from OpenStreetMaps back to the census block group level. By getting a count for the number of assumed residential buildings in every census block group, I can relate an estimated population and number of vehicles to the individual shape files for each building. I will use these buildings as an approximation for vehicle location i.e. if a building shape file overlaps with a flood prediction, I translate that effect 1:1 to its associated vehicle count. A graphical representation of this concept is included below.

Defining Annual Average Loss

The idea is that I am using the floorplan of residential buildings as a proxy for where cars might be left and be damaged by a flood. I am including a couple of caveats here: one, is that some percentage of vehicles may be safely moved with advance warning of a large storm event. I estimate that percentage to be about 40%. From Kelly Blue Book, I also get an average value of a sedan to be $30,281. These factors are used along with the earlier percent damage relationship to flood depth. From here I can get a reasonable estimate vehicle cost of every flood scenario to each building.

What is potentially also interesting however, is how the risk of each scenario over several decades materializes into an average annual vehicle flood cost. So I will choose to inspect what this looks like from 2020 to 2050 in 10 year increments. Using RCP4.5 models to predict the likelihood of each flood scenario, I can combine the expected damages and costs of every scenario in different years. Two variables that I’ll take into account that change year to year is the probability of a given SLR (it is far more likely to have 25cm of SLR in 2050 than in 2020), and the number of vehicles in Foster City. The SLR probability comes from climate models that I found here(https://github.com/bobkopp/LocalizeSL), and the vehicle predictions come from EMFAC projected ownership in San Mateo County. I looked at the percent changes in the county and multiplied my count for Foster City by the same numbers. The following plot shows how this annualized average loss (AAL) is spread across the region and how it varies in different years. In 2050 shows a far higher AAL than 2020 because the whole of Foster City is at much greater risk in the case that sea levels rise and breach current earth barriers.

This map gives a sense of how extreme the cost of this jump in flood risk is at the individual level in Foster City as sea level rises become more likely. Keeping in mind this is just an estimate for vehicles, the AAL becomes ludicrously expensive for each building. This modeling does not take into account the investments in new protective infrastructure that is already underway in Foster City. Seeing these sorts of predictions, it makes sense why a multimillion dollar investment in sea walls would seem a wise investment to limit future costs to residents. However, this sea wall will change these flood predictions, but it is very difficult to predict how this new intervention will affect other cities around the bay. Given the hydrological tide patterns in south bay, this sea wall will almost certainly exacerbate the extremity of future tides at Eden Landing and other coastside regions. This sort of damage prediction is alarming, but must be headed with a consciousness to the interconnectedness of physical systems.